Okay. For the first session, I would like to ask our CEO, Johnny Shen, to start the meeting with a very brief company introduction. Johnny.
Good afternoon, ladies and gentlemen. This is Johnny Shen, President and CEO for Alchip Technologies. Once again, thank you for attending the Q3 Investor Conference Meeting. We appreciate the opportunity to share our Q3 result, and also provide the guidance for future business. In case some of you are not familiar with Alchip, allow me to make a quick company update and identification. Our company is founded in 2003. IPO in Taiwan Stock Exchange, 2014. Our current headcount is 485 people. Majority of our employees are engineers. Since we founded the company, we've been focused on ASIC business. Right now, we have successfully taped out more than 470 designs, all in leading-edge technology. Revenue last year, we are achieving $240 million.
This year, we improve quite a bit. The first three quarters will be over TWD 283 million already. Our current capacity, we can support 30 designs concurrently. 80% of our business is focused on leading-edge technology, mainly for HPC and AI. We are one of the VCA members for TSMC, which can distribute TSMC's technology worldwide. Next page, please. Okay. Overall, we can see we had another outstanding quarter with a reasonable number, almost as good as previous quarters. Q3 actual number really outperform our planned number, which we made earlier this year. All detailed analysis and comparison will be provided by Daniel in a later section. Q3 achievement is not only in numbers.
Most importantly, we have established few new business from so-called dream customer in different application, yeah, including cell phone and automotive, et cetera. We had multiple tape out and design win, mainly in 6 nm, 5 nm or 7 nm. Of course, HPC and AI, yeah, including CPU and GPU, 2.5D CoWoS related business demand remain very strong from existing repeated customer, also from newcomers. As for the mass production forecast, we receive an incredible high the volume from key customers. Number is too large to digest. Honestly, if we can hit, yeah, we can easily hit another home run if we can get the support from supplier to fulfill just 50% of their forecast. Yeah, honestly, the MP forecast we receive is incredible high. There are two major challenge for our business. Number one is the die resource.
Nowadays, recruiting or retaining engineers is the most critical challenge for the company. I want to thank you very much for the share price support. It makes our senior turnover rate almost zero. Another challenge is the capacity allocation, not only for wafer, but also for substrate. Almost all HPC and AI suffer ABF substrate shortage. Fortunately, most of our suppliers are buying our story and future, and the future potential. Since we are representing more than 30 customers, working with TSMC helps suppliers to diversify their business concentration without too much effort. Lastly, I will make a quick update for Phytium status. Since the incident happened 8 April , we are working with customer, U.S. authorities, major supplier, and also government authorities closely, trying to address all parties concerned. The situation so far, so good toward a positive direction.
At this moment, we have a consensus agreement from all parties. We will no longer disclose the detailed progress for each of our Feiteng product lines, unless there's any rumor or showstopper happen, which need us to clarify. Yeah, we are looking forward to resolve all issue soon. Yeah. Above is the update from me. Thank you.
Okay. Starting from this session, the number is pretty straightforward. The detailed PNL numbers, we already upload to the MOPS. For third quarter, the revenue is $91.9 million, which is 38% year-on-year growth, but 5.7% quarter-on-quarter decline. For the operating income in third quarter is $15.3 million, which is 84% year-on-year growth and 12.4% quarter-on-quarter decline. For net income level, because of the lower operating expense and a little bit lower income tax, the net income for third quarter alone is $12.9 million, which is 57.9% year-on-year, and a 7% decline quarter-on-quarter.
In terms of EPS, the third quarter earnings per share is TWD 5.12. This page is for the breakdown by application for third quarter. You can see the HPC related revenue still dominate our total revenue, which accounts for 81% of our total revenue. For the first to third quarter this year, HPC accounts for 87% of our total revenue. For the consumer part, consumer accounts for only 4% of our total revenue this year.
For the revenue mix by process node, for this quarter, as you can see that, because we already hinted that, our major revenue contributor in the first half, which is our second-largest China customer, their first 7-nanometer chip shipment was tapering off in third quarter. That's a reason for the 7-nanometer revenue contribution to decline quarter-over-quarter to 44%. But for some other projects, they received quite strong demand from our customer. For the third quarter, 7-nanometer or even 6-nanometer, 5-nanometer, plus the 16 and the 12 nanometer, combine these process nodes accounts for 86% of our total revenue.
For this year until the end of third quarter, the leading-edge technology node, 7 nm, 16 nm, 28 nm, accounts for almost 95% of our total revenue. In terms of the position within the leading-edge technology node market, I would say Alchip is definitely the leader in this field. For the revenue breakdown by region, as you can see, China remains our biggest market for our revenue. In third quarter, revenue to China customer account for 75% of our total revenue. Japan accounts for 11% and Taiwan 8%, and the others accounts for 11%.
For this year, until the end of third quarter, China market revenue to China market accounts for 76% of our total revenue, while the other region accounts for 15%. We expect the other, which include the North America and the Middle East, the percentage surge significantly in 2022, when our 7 nm AI projects began to contribute to our sales. For the third quarter operation, there are mixed reason for better than expected third quarter results. First of all, increasing orders from China major customer and the NRE project inflow is really strong.
The new projects with high volume projection, we won several of them from China service provider, which we are pretty excited that this project is a major win for us to get into the China service provider supply chain. We expect the production revenue to contribute a lot starting from next year. For the operating income, which is TWD 50.2 million for the third quarter, is 84% year-on-year growth. You may see that for third quarter, gross margin went down slightly. That's because the NRE percentage in third quarter is a little bit lower than comparing to the second quarter.
For the production revenue, among the production revenue, the exposure to some high-margin projects is a little bit lower, comparing to the second quarter. That's the major reasons for the gross margin went down. For the operating expense, which is about TWD 1 million lower than our second quarter operating expense. The main reason for the lower operating expense is we do less bonus provision for the third quarter. For the fourth quarter outlook, again, the industry-wide demand feels strong. I believe that most of you even know this better than I do. The whole demand within the semiconductor industry is really good. The design demand within the leading-edge technology nodes remain robust.
Many projects out there waiting for design resource. That's the reason why we always mention that the current design resource is in shortage. We really try our best effort to recruit people to extend our design resource in order to serve more customers. The HPC demand keeps its momentum as increasing applications in the participants. This industry-wide concept, I also guess, all you already knew it a lot, that because of AI, because of there are many, many new concept coming out of market. So along with that, there are a lot of new application emerges within the HPC or let's say AI field. Of course, there are more and more participants with more and more design opportunities out there, which are all contribute our operate...
Operating performance. Last thing is the AI chip. Our 7-nanometer AI chip shipment to increase its contribution going forward. The first 7-nanometer AI chip to U.S. service provider customer, we tape out in the quarter. We expect the mass production shipment to start in fourth quarter. After ramping up, we think next year revenue contribution from this project will be huge. Our CEO mentioned that the customer is keeping on adding to the forecast. The headache right now is we need to find suppliers to support. It's not an easy task to do. We are trying our best to do that.
Another AI chip to the U.S. chipmaker is also expected to start its contribution in fourth quarter this year. Another one to the same U.S. chip maker, which is a training chip, expected to start in the first quarter next year. We also believe the shipment volume could be considerable. All these three AI chips to U.S. customers are expected to be one of the growth drivers to our 2022 revenue performance. I think that's the end of the company presentation session, and we will quickly go into the Q&A session. As mentioned, please tell me your name and the company you're working for and. Okay, let's go into the Q&A session.
Haas, please, you can unmute your microphone.
Okay. Hi, Johnny. Hi, Daniel. This is Haas from Credit Suisse.
Hello.
Hello, can you hear me?
I can hear you. Go ahead. Thank you.
Okay. Congratulations on the good results, and thanks for taking my questions. My first question is regarding your 2022 business outlook. As we get closer to the year-end, could you provide us your guidance for fourth quarter and also for annual target for the sales growth and also margins outlook, by turnkey and also by NRE, for next year, if we don't consider Amphion contribution? Thank you.
Okay. I would say we don't have a numerical guidance for next year, and we are not allowed to give it because of the TSE is requesting us not to deliver any numerical guidance to the market. For the picture, I guess, Johnny can talk a little bit about it for our 2022. We do believe as long as we can resolve the supply chain issue, 2022 will be another great year for Alchip. For the fourth quarter, since the AI chip to U.S. service provider the shipment is just going to kicking off. The contribution in the fourth quarter is not that significant compared to the forecast last year.
In fourth quarter, we still believe it's an NRE heavy quarter. We expect the total revenue will be like flat to mildly quote-unquote growth for our fourth quarter. We also expect the margin, the gross margin will remain at the same or a little bit higher level comparing to the third quarter. For 2022, as I mentioned, there are many major projects. 7 nm projects will go into production phase. According to the forecast that we received right now, the silicon volume could be incredible. Yeah, again, the substrate supply is a headache right now. Johnny, are you going to go to? Yeah.
Sure. Haas, to answer your question, 2022, I think we have quite confidence. First of all, NRE. As you know, the demand is very strong. Even the existing customer, all existing customer has their next generation repeated project. Many newcomer, not only in China but worldwide. In terms of NRE, the only challenge is our design capacity, is our design resource. In terms of number, as you know, the order supply increase. The NRE. That definitely will reflect on our revenue. Mask price increase, IP price increase, NRE EDA tool increase, is, I think it's all adding boost to our revenue. Also since our design capacity is very limited, and we also increase quite a bit for our own design margin.
Yeah, that's the overall situation. Also, technology migration. This year we are focused on 7 nm, mainly 7 nm. Few 5 nm and few 16 nm. Next year, our concentration will be 5 nm, and we expect some 3 nm, some 4 nm, and 6 nm. Each of the generation NRE increase is quite significant. That's the NRE part. For the mass production, like I mentioned before, recently we received incredible high volume forecast. Yeah, honestly, and that is too big for us to digest. Yeah, like I mentioned before, we could easily hit another home run if we can get the support from supplier just to fulfill 50% of the current demand. That would be incredible already. This one, MP revenue still have a certain uncertainty, yeah, because of substrate shortage. Yeah, hopefully we have a breakthrough discussion with our suppliers soon.
Haas, does that answer your question?
Okay, that's very helpful. Before discussing on the capacity constraint, just a quick follow-up. Could you share your mix for NRE and also turnkey for 3Q and 4Q? Also your expectation for next year. Thank you.
Mm-hmm.
Okay, so the new NRE coming in, I would say, as I mentioned in the slides, we won an AI-related project, but more to the end user side, project from Chinese service provider, which we expect to have mass production contribution, most likely, in the second half of next year, which is a very high volume project. For other projects, there are too many. I'll give you several examples. We already won a project from Chinese AI player, which is also considered a 6 nm project, I think. And the...
Because of the end application is handset related, we do believe the production opportunity and production volume could be good. There are too many. We are going to close a project which is also which is for the mobile devices with one U.S. customer, which is also considered a high volume project. As I mentioned, we are handling Image Signal Processor project which Taiwan is vendor and the end application is handset with China handset vendors. All in all, as we keep on emphasizing that NRE is not a problem to us, I would say how to choose the projects wisely is the top priority we are facing right now.
Okay, thank you. That's very helpful. My second question is about the turnkey business and the overall capacity constraints. You mentioned you still have a significant gap to meet your customers' demand on 7 nm and also 5 nm. Could you try to quantify how much demand at this stage you are unable to fulfill? How much more foundry capacity do you expect to secure for next year versus this year? Are you worried about the potential overbooking from your customers, especially during the time of the foundry supply tightness? Thank you.
Honestly, we don't worry about it because we do see very strong demand for the HPC related projects. In terms of the percentage, I would say this way. For the wafer capacity, we don't worry too much because we have a very good customer portfolio. Customer and Alchip will work together to try to get the capacity from TSMC. For substrate, because as you may know that, the current substrate capacity for next year is almost dominated by the big players like Intel, like Nvidia, like AMD. Trying to get some capacity support from the substrate vendor is more difficult than getting wafers. For wafers, we don't worry too much. For substrates, we are working very hard.
Because the customer is keeping on adding forecast numbers to us, for the percentage right now, as Johnny mentioned, if we can provide 50% of the demand, we will be in a very good shape next year.
Yeah. Haas, to add on to your question. Originally, I think our fulfill rate is reasonable, but somehow recently, customer keep increasing their forecast. Once one particular customer original demand is already high. But somehow, within half a year, they quadruple it. I think it's almost impossible to fulfill their demand, their forecast. In terms of substrate, we've been talking to many vendors. In fact, if a customer come to them individually, they deny all the meeting, but they still work on Alchip. The reason they want to work with us, because we are representing more than 30 customers. They also need to diversify their business concentration.
I think we can say we get a great support, but of course, not 100% because all the capacity is booked by the Tier 1 customer Daniel mentioned before.
Okay.
Mm-hmm.
Sorry.
Oh, go ahead, go ahead.
Okay. Also, just a quick follow-up on how much your foundry partners price hike do you think will help your turnkey business growth in 2022? And will you be able to pass on all those costs or even add a little premium to your customers to keep or to raise your turnkey margins? Thank you.
Mm-hmm. Okay, so-
Okay. Johnny, please.
All right. Okay. As you know, the application we are focused on is HPC and AI. Fortunately, those fields are price insensitive, much more price insensitive compared to consumer. Fortunately, all our customers buy into TSMC's pricing increases, so do all the suppliers, who have been successfully transferring all the pricing increases to our customers. They are willing to pay. Most of them are willing to pay. Because of design capacity shortage, they are fighting for our design resource. They also expect certain NRE increase as well. To answer your question, we don't have too much trouble to transfer all the pricing increase to our end customer. Daniel, do you want to add more?
Yeah. Similar to Johnny. Actually, I talk this issue with the investors many times that for the majority of our production, pricing is a cost-plus margin. For this part, we have no problem to pass the increasing cost of wafer from wafer to our customer. For some very limited customer product, maybe there will be a mild impact to the cost. We may have to share some, but since the consumer revenue accounts for very low percentage of our total revenue. All in all, we don't see the wafer price appreciation is a threat to our operation.
Okay. Is there a way to quantify the price hike from the foundry partners that will help your turnkey business? Is there a way to quantify the impacts? Thank you.
No, we didn't do the calculation.
Okay, thank you. My next question would be on Feiteng. You made some good progress, as you just mentioned, but should we expect shipment to start pretty soon? Or, at this stage, there's no update yet?
No comment.
Yeah, no comment. Yeah, we can only say so far so good and no comment. That's a agreement from all parties. Yeah.
Please understand. I think, for the Feiteng thing, no comment is the best strategy we can do right now.
Mm-hmm.
It is good for our customer, and it is good for our supplier, and of course, it is good for LHB itself. Forgive us. That's the agreement among all the parties that we don't comment on the details of Phytium's project anymore.
Okay. Thank you. My last question before I shift back to the queue is that I would like to check on the R&D engineer sourcing. You mentioned you have 470 employees as of 3Q, but you originally had a target to expand it to 500-600 employees by the year-end.
Mm-hmm.
Could you discuss if any change on the target, and do you think it will be a bottleneck for you to take more projects? Thank you.
Participants, please mute your speaker. Otherwise your background noise will go through the meeting.
Okay, Haas, to answer your question, the plan changed a little bit. By the end of this year, we still have a confidence to be over 500 for sure. Yeah, because of the in China and next year, starting from next year, we do have a plan to increase another 200 people. Actually, the offer already made. They supposed to join by the first portion will be March. Second portion will be July. Yeah. Fortunately, the senior people turnover rate, as I mentioned, is almost zero. Yeah. Because most of them are compensated from the stock share price increase would become majority of their income source. But junior people, to retain the junior people is still quite challenged.
Yeah, because of the junior people, they only have a salary. They don't have a stock. Next this year we can hit more than 500 for sure. Next year our internal target is 750 people. Right now, more than 100 offers already made. Hopefully they will sign back soon.
Okay, most of the new headcount will be in China or you are diversifying it?
In China, Japan, Taiwan. You can say majority still in China. Yes, you are right. We're doing hiring worldwide, not only in China.
Okay. That is very clear. Thank you so much, Johnny and Daniel. I will be back in the queue.
Thank you, Haas.
Okay. Szeho , please, you can unmute your speaker. Yep.
Thank you, Johnny, and congratulations. May I ask two questions, basically. The first one, just want to check your brain. How do you think about the thickness of your, U.S. hyperscaler business, when they move from generation to generation? Basically, you are still seeing very strong visibility into 2022. I'm not sure if there will be any risk for them to move to other design service houses or you're already working on their future projects.
Okay. Let me answer your question first. For U.S., for the U.S. customers, of course, we want to do generation by generation.
Mm-hmm.
on projects. We do have confidence to do so.
Mm-hmm.
In the end, it's up to the customer to do the vendor selection. The situation right now is almost all the customers, especially for the service providers, are trying to do multiple projects at the same time. You may see the news that Facebook is aggressively doing their chips.
Mm-hmm.
Alibaba has their own CPUs and, not to mention AWS has many new projects. Of course, we cannot grab them all.
Mm-hmm.
To us, the concept is, there are increasing design opportunities from those hyperscaler players. Because of our position within the leading-edge technology nodes, we are pretty confident we can win some of them. Actually, we already won some of them. But for the future, again, the decision is made by our customers. We cannot guarantee to our investors that no problem, we can win. That's not the thing we are going to promise.
Yes. You're quite confident, right? Yeah.
Uh-
Given your competitive position.
Yes. Within the leading-edge technology, actually, we almost have everything. We have a good track record. We have excellent engineering. The customer satisfaction is high for us for their previous projects. Actually, I think our position is good.
Sounds great. My second question, going back to your slide. In one slide, you mentioned that you're working on 30 projects simultaneously, right?
Mm-hmm.
Let's say going into 2022, when a lot more projects going through the final geometries, like the 5 nm, 3 nm. Would the number of projects be similar even though you keep adding more accounts? Yeah, because each project would suck up more engineering resources. I'm not sure if that's the correct reasoning.
Yeah.
Szeho , you are absolutely right. Yes, I think if you pay attention for our design capacity, will be increased insignificantly because technology migration. Just to share one information. For example, to doing 60 nm design, in usual case, we only need 20 people. Now, if you're doing a seven or five, we need 40 or even 50 people. So increase the number of headcount, eventually we're doing a five or three, need more resource. So focus on number for design is not our intention. Like Daniel mentioned before, how to select the right project will be our highest priority. The new design using more people and a lot of machines. So number of design I don't expect will be increased too much, maybe fractional.
In terms of technology, in terms of NRE, yeah, we have a high confidence to have a very reasonable growth.
Okay, sounds good. More headcounts required for each project, but the design cycle time should be similar, right? Even if we move to the smaller nodes.
Yes. I think the design cycle, if we use more people, that will be more parallel effort. We try to make the design cycle the same. The design cycle I think dictated by many factors, machine performance and also customer experience. Based on our track record, if it's a startup company, usually the design cycle will be 20% longer. If the same customer doing the repeated business, we can easily shorten the design schedule by 20%-30%. That's why one of our priorities is a repeated customer instead of a new startup company. That will save a lot of TAT. They can increase our capacity.
I see. Yeah. Sorry, the very last question, at this point in time, we still don't have any plan to build our own IP library, right? Even for some select applications like advanced packaging, you still don't have the plan.
Yeah. For IP, yes, we still try to remain independent position. Yeah. Like I mentioned before, leading-edge technology design, HPC, AI, using many complicated IP and different technology nodes and different purposes. At this stage, it's very difficult for us to design it all. Instead of compete with a partner, we heavily rely on partner, and we can work together on the IP side. Yeah, so far so good. Yeah, because it's almost impossible for us to prepare all the IP. Yeah, right now, I don't think IP is a showstopper for our business.
I see. All right. Okay, thank you very much. That's very useful. Thank you.
Thank you.
Charlie, please unmute your speaker.
Thanks, Daniel. Hi, Johnny. Just some follow-up question to previous question. First of all, on the IP, I'm not sure if you still see the entry barrier for the CoWoS implementation. Do you think your, for example, the production yield is in line with industry peer? Or do you see there's actually no difference anymore for CoWoS implementation?
Okay. Let me try to clarify. This is not the first time I've heard somebody compare or complain about our CoWoS solution. In fact, most of our design, more than 10, right now it's three to four already in volume production. Any CPU, GPU, AI training related application are 2.5D CoWoS application. We are winning a lot of them already in volume production. In fact, based on our forecast next year, CoWoS related design MP volume will dominate our entire MP business. I don't think there's any showstopper for CoWoS related business for us.
In that case, is it the right way to ask, you know, how much CoWoS will contribute to your revenue or compared to the previous kind of substrate base, you know, that's a packaging. What would benefit you more?
Charlie, we don't have a percentage right now. I'm lazy. I didn't do a statistic thing. As Johnny mentioned, I think, let's say this way, many investors may think that our peers has their own in-house CoWoS, which implies that we don't have the CoWoS ability. That's totally wrong. I think, as Johnny mentioned, one of the biggest revenue contribution next year is using CoWoS. I believe in terms of the production revenue contribution, CoWoS will be CoWoS related revenue will be pretty dominant to our overall total production revenue next year.
I see. Thank you. Also, Daniel, you just bring a great point, right? Meaning, lots of as an internet company or data center customers, they do their own projects as well, right? Like, Graviton from AWS or Alibaba T-Head, right? As a CPU.
Mm-hmm.
I know you cannot address all those demand, right? But from your observation, those are big guys. What kind of projects they want to outsource to you or your peers, and what kind of projects they prefer to do themselves?
Yeah. It is also a puzzle to us that we know, for example, AWS did their Graviton and Alibaba do their own CPU project. Based on my knowledge, I don't know who did the backend design for AWS. For Alibaba, I know that they did it on their own. The whole process is very difficult and, let's say this way, as long as they find out that doing the backend design is not that efficient on their own, they may consider to outsource. And/or when they are trying to do multiple projects, but their internal design resource is limited, they will consider to outsource.
For us, we don't know what's the category for them to select which projects is going to outsource and which project they are going to do it internally. To us, no matter how it is a good news, the design opportunity out there, the number of design opportunity is increasing. Because of our position, we always have chance to win some of them.
Thank you.
Johnny, do you want to add some color?
Okay, sure. Like Daniel mentioned before, so many design opportunities out there, no matter how hard we work, it's impossible to grab it all. Some of our efforts, some of our potential customers were doing by themselves or share some of the design to our competitors. I think that's a good thing. It's not a bad thing for the companies. We have a competency to maintain a competitive position. We always believe and also share with the customer, each of our customers, they only do one project or two projects every year. We're doing 30. The knowledge accumulation, the risk management, we are doing much better and more efficient because we have more experience.
In fact, the turnaround time, if we're finalizing, you can save a few months at least, because we did it before. Also, if they, from an investment point of view, doing one design, they need to invest so many EDA tools and servers. Yeah, for us, it's highly justified because we have many customers doing all together. I think based on this trend, the track record and design experience, we can always maintain a competitive position. Yeah, much better than customers doing by their own or most compared to most of our competitors.
Okay, thanks. Lastly, yeah, 5 nm, right? I mean, I'm not sure if last quarter or two quarters ago you mentioned,
Mm-hmm.
There are 2 5 nm NRE.
Mm-hmm.
What was the update? Any, you know, more design wins for 5 nm, even 3 nm you just mentioned?
We do have a few. I don't recall which one, but in terms of 5 nm, we have many design wins recently. For 3 nm, not yet. We do have some test chip opportunity. We helping one of the customer to do the test chip, and also we are doing our own test chip, proving few of the internal IPs by our own. Daniel, do you wanna add more?
Okay. Johnny, I think you are referring to the 3 nm, the 5 nm test chip to North American service provider. I think they are preparing for the full mask design. We don't win it yet. As you know, this service provider has many projects out there, and we are engaging in discussion with them for some of them. As long as we have good news, I will let the investor know. So far, they are still preparing the, I would say pre-project kickoff preparation right now.
Yeah. For those, the project, they can enter mass production. You know, can you-
Okay, by the way.
Give us some... Yeah. Mm-hmm.
By the way, the first 5 nm full mask project, we are going to tape out probably in the first half next year.
What application for region?
AI to the Middle East customer.
Okay. Is that the only one in 2022 or any more tape outs?
I am not so sure about if we can have multiple tape outs next year. We are winning several five nanometer projects and the design is going to kick off sequentially. For the design time, a long time for five nanometer, I am not so sure about if we can tape it out by the end of next year.
Okay. That is fair.
Mm-hmm.
Thank you. That's all my question. Thank you.
Okay.
Thank you, Charlie.
Jeffrey, please, you can unmute.
Great. Okay, Johnny, Daniel, can you hear me okay?
Yes.
Yes.
Yeah. The first question, you know, the fundraising you did early this year, one of the key reasons for that was to prepay foundries. Has that helped you get more capacity secured for next year? How much have you prepaid?
Actually, for this question, for the whole scheme with TSMC, then, the schemes like, we put money for deposit and we get credit from TSMC. At this moment, actually, our POs to TSMC already exceed the credit they give us. We need to pay in cash. We are negotiating with TSMC to raise the credit line for us.
Mm-hmm.
On the other hand, the preparation for the funding, one of the major reasons is to let's say for the 5 nm tape out or even future 3 nm, the mask cost is huge. If we cannot lift up our credit line by TSMC, we need to pay it in full. That's the reason. So far, so good. Our cash position is still very good.
Mm-hmm.
We are preparing for next year. On the other hand, for substrate, we do need to pay in advance to secure some capacity for next year and even 2023. The funding in the early days of this year do help.
Okay, great. A follow-up question. You know, if I read correctly on your percent of sales by technology node?
Mm-hmm.
Over 40% in third quarter was 40 nm.
40?
Did I read that incorrectly?
You mean?
Third quarter sales by technology node.
No. Third quarter, 7 nm or 5 nm accounts for 44% of our total sales. The 16 nm and the 12 nm accounted for 42% of our total sales.
Okay.
Yeah. For 28, it's 9%. For 40, it's 5%.
Okay.
Mm-hmm.
You made an analogy earlier to hitting a home run. If I did my math, is it double 20%+ revenue growth, triple 30%, and a home run is 40% revenue growth for next year? Or what is how do I-
Sorry, I cannot provide you the numerical guidance.
Yeah.
We already warned by the TSE many, many times. They are warning that if we give the sales growth guidance in numbers-
Yeah.
We will be forced to do the financial forecast.
Okay. Understood. The last sensitive question I have, you mentioned all parties to the Feiteng kind of tax agreement. That includes U.S. government, I assume.
No. Actually, we don't have the position to talk directly with the U.S. government.
Mm-hmm.
For our customer, Alchip, and our supplier, mainly TSMC, we do have discussion about this issue.
Mm-hmm. Okay.
Yeah. It's good for everybody that we don't disclose the details of the Feiteng projects.
Okay. Great. Understood. Thank you.
Mm-hmm.
Thank you, Jeffrey.
Mm-hmm. No problem. Charlie, do you have more questions?
Oh, yeah. I just noticed that your OPEX in third quarter drove the loss. You said that it's because there's a bonus provision. I'm not sure if there is also other factors. For example, I believe the legal or lawyer expense should be quite intense in the recent quarters. What
Yeah.
Yeah. What is the kind of the new norm for your OPEX in the coming quarters?
I think you can say that TWD 15 million per quarter will be a new norm to our OPEX. For third quarter, I think the major reason is relatively lower bonus provision. Of course, for some legal fees or some others, there are some savings. That's a mixed result for many items. If I want to add our opinion on the norm of our operating expense per quarter, I would say like TWD 15 million. TWD 15 million or TWD 16 million is a good guess.
Okay. Thanks. Yeah, actually, just to clarify, you said you can only fulfill half of the demand. Did you refer to the turnkey demand or NRE demand or both?
Only turnkey demand. The NRE, actually, our fulfill rate depends on how you calculate. If it's for the project we try to win, I think our fulfill rate is very high. We can say it's 80%-90%. But in comparison to all the potential business, I think we don't have a position to grab one half. So many opportunity out there.
Okay.
Charlie, let's say it this way. The industry demand is really, really strong. For example, I think in previous institutional investor meeting, because of the Phytium incident, we told the investors that our target for this year is 20%, which means TWD 300 million for the whole year. From our previous target prior to the Phytium incident. Prior to Phytium incident, yearly revenue target is about TWD 360 million-TWD 370 million. I guess most of you already knew it. Because of the industry-wide demand so strong, you may see based on our guidance, fourth quarter, our full year revenue actually exceeds our guidance prior to Phytium incident. From the numbers, you can...
I hope you can feel a little bit about the strong demand from the industry.
Okay. Yeah. Just out of curiosity, right? For those, the unfulfilled NRE projects, do you think that those are perishable or, you know, go to your industry peers? I'm not sure how real are those requests.
We don't know the situation of our industry peers, but honestly, I do think almost all our industry peers are in a very good shape. I read some report that say Global Unichip, their project is also full. There are even some projects bidding to like VeriSilicon, which means the demand is real. It's not the air projects. Those projects, they do need backend design support and maybe the turnkey help, but they cannot find resource. I think that's the current status for the whole industry.
Okay. I see. Thanks. It's super helpful. Thank you. All right. Thank you.
Okay. Because it's already 3:30 P.M., can we take the last question. Okay? Is there any questions? Okay, Jeffrey, please.
Yeah. Daniel, did you give a breakdown for NRE versus turnkey in the third quarter?
Okay. Sure. Wait for me a moment.
The follow-up question is fourth quarter outlook for that.
For the third quarter, NRE accounts for about low 40s% of our total revenue.
Any major change you expect in the fourth quarter or, like you mentioned, NRE seems like it's growing faster?
For fourth quarter, we expect NRE percentage will be higher, like high 40s%.
Okay. Great. Thank you very much.
Thank you, Jeffrey.
Okay. If that's the case, thank you. Thank you all for joining our institutional investor meeting, and have a good weekend. Thank you.
Thank you very much. Thank you for your support. Thank you very much. Thank you.